This is excellent, I've always wanted a simple db library that makes defining a schema upfront optional. After all a schema is a constraint that can be set in

This is excellent, I've always wanted a simple db library that makes defining a schema upfront optional. After all a schema is a constraint that can be set in stone later on, just like adding indices for optimisation purposes. Few questions: - what's the performance like? is there any overhead to alchemy, eg comparing schema every time you do an insert? - no way to specify primary keys as an alternative to the auto-generated id column? - no table.remove()? - what about inserting more complicated data structures, eg a dict with a nested dict or a list? would be great if those were serialized auto-magically into a blob type (or used to create another table with a foreign key?) - would be nice to be able to freeze a schema with table.freeze() for example: from then on new columns don't get created automatically, or get stored in an extra blob column (this is a very common scenario for python devs where non-indexable columns just get stuck in a blob) - let me optionally define a schema and specify defaults with table.schema(name='', price=0.0') - would love to see table.ensureIndex('column', 'unique') similar to mongo for quickly creating indices - db = Dataset() should do dataset.create('sqlite:///:memory:') for me - would be nice to have that as the default connector, so that Dataset() acts as LINQ for Python by default - dataset.freeze is nice but I'd rather have dataset.export() & dataset.import() letting me easily copy rows from one db to another (after inspection for example) Thanks for creating this!